Participants visited the laboratory on 5 separate days (1 control trial, two 15-min experimental trials, and two 43-min experimental trials). At each visit, they completed an exercise bout on a cycle ergometer (LC6 Novo, Monark Exercise, Vansbro, Sweden).

The first visit was a control trial; participants completed a graded exercise test (GXT) to measure maximal HR (HR_max) and V ˙ O_2max in a temperate environment [22.6°C ± 0.6°C, 36.6% ± 5.8% relative humidity (RH)]. The remaining 4 experimental trials were completed in a counterbalanced order and a hot environment (35.1°C ± 0.3°C, 40% ± 4% RH). Counterbalanced treatment orders were randomly assigned to participants. Each experimental trial consisted of an 8-min warm-up at 70% HR_max or an RPE of 12, followed by 1 (15_HR and 15_RPE) or 5 (43_HR and 43_RPE) rounds of HIIT using HR or RPE to prescribe the intensity. One round of HIIT consisted of 4 min at 90% HR_max or RPE of 17 and 3 min at 70% HR_max or RPE of 12 (Figure 1). If 70% HR_max or an RPE of 12 could not be achieved during the rest intervals (because of thermal and cardiovascular strain), participants cycled at 30 W with a cadence ≥30 rev⋅min⁻¹; 30 W was selected to ensure participants were not recovering passively and were still cycling against resistance. Upon the completion of each experimental trial, without cessation of exercise, participants immediately began a GXT, performed at approximately half of the maximal power output observed during the control trial, to measure V ˙ O_2max. Because cardiovascular drift typically occurs after 10–15 min of exercise, necessitating work rate adjustments to maintain THR or RPE, the purpose of the separate 15- and 43-min trials was to evaluate V ˙ O_2max before (15-min trials) and after (43-min trials) work rate adjustments had been made in order to maintain the prescribed intensities. Additionally, the 15-min trials were needed because it is not feasible to measure V ˙ O_2max at 15 min and 43 min within the same trial.

An a priori power analysis (G*power 3.1.9.6) revealed a sample size of 7 would be sufficient to detect a 25-W difference between the change score in power output from the first to the final work interval in 43_HR versus 43_RPE, assuming α = 0.05 and power ≈0.80 (Faul et al., 2009; Faul et al., 2007). Eight healthy adults (4 men and 4 women; 18–38 y) free of disease participated. Seven were recreationally active as defined by the American College of Sports Medicine (2022) (i.e., exercising at a moderate intensity aerobically ≥30 min per day, ≥3 times per week, for the past ≥3 months) and 1 male was a competitive endurance athlete. Physical characteristics of participants were age (mean ± SD) = 25 ± 7 y, body mass = 74.1 ± 8.3 kg, height = 181 ± 10 cm, percent body fat = 21.4% ± 8.4%, V ˙ O_2max = 3.2 ± 1.2 L⋅min^–1, HR_max = 185 ± 5 b⋅min^–1, 70% HR_max = 130 ± 5 b⋅min^–1, 90% HR_max = 166 ± 7 b⋅min^–1.

Women with a regular menstrual cycle lasting 21–35 days were included (Elliott-Sale et al., 2021). They were asked to self-report the first and last day of previous menses and contraceptive use for data analysis and scheduling. All experimental trials were scheduled during the same phase of their menstrual cycle (luteal phase or follicular phase), although the specific phase was not expected to affect study outcomes (Stone et al., 2021). Two of the 4 women completed the experimental trials in the luteal phase of their menstrual cycle. Although phase of menstrual cycle was not confirmed via hormonal assay, based on cycle reporting and rectal temperature (T_re) it is likely 1 woman completed the 43-min trials in the follicular phase and the remaining trials in the luteal phase. One woman who was using an oral contraceptive [norgestimate (0.25 mg) and ethinyl estradiol (0.035 mg)] was tested in her follicular phase.

For each of the 5 trials, participants were instructed to abstain from consuming alcohol or participating in strenuous exercise during the 24 h before testing. Additionally, participants were asked to report to the laboratory well rested, euhydrated, and having refrained from ingesting non-prescription drugs and caffeine on the day of testing. Pre-testing instructions have been successfully used by our lab previously and adherence was confirmed using a 24-h history questionnaire (Yoder et al., 2023; Mulholland et al., 2023).

Upon arrival, participants provided a urine sample that was analyzed for urine specific gravity (U_SG) using a digital refractometer (ATAGO PAL-10S digital refractometer, Tokyo, Japan). U_SG had to be ≤1.020 for a participant to be considered adequately hydrated (Sawka et al., 2007). Participants whose U_SG values were >1.020 were given water to ingest for 20–30 min and then reevaluated. Then participants dressed in padded cycling shorts and a mesh tank top and were equipped with chest strap HR monitor (H10, Polar Electro, Kempele, Finland) that paired with a smartphone application (Polar Beat, version 3.5.0, Polar Electro, Kempele, Finland). Prior to beginning exercise, the Borg 6–20 RPE scale was explained using standardized instructions (Borg and Noble, 1974).

A minimum of 24 h separated control trials from subsequent experimental trials and at least 48 h separated experimental trials from one another. All trials for a given participant took place over ≤8 weeks and each was completed at a similar time of day to control for fluctuations in core body temperature associated with circadian rhythm (Moore-Ede et al., 1983).

At the first visit, participants completed a questionnaire about their readiness to participate in exercise and a general health history form. Next, height was measured using a stadiometer (SECA 213, Seca Ltd., Hamburg, Germany) and body mass was measured with a digital scale (Tanita WB-800S, Tanita Corp., Tokyo Japan). Body fat percentage was calculated from the sum of 3 skinfolds (Jackson and Pollock, 1985).

Participants then began a self-selected warm-up for 5–10 min on the cycle ergometer. Next, the GXT started and every 2 min the power output on the cycle ergometer was increased by 25 W until volitional exhaustion was reached or pedal cadence fell below 30 rev⋅min^–1. V ˙ O₂ was measured continuously using open circuit spirometry (Parvo Medics Metabolic Measurement System, model TrueOne 2400, Salt Lake City, UT, United States). V ˙ O_2max was considered as the average of the highest 2 consecutive 30-s values. During the GXT, HR was measured continuously using a smartphone application (Polar Beat, version 3.5.0, Polar Electro, Kempele, Finland) that was integrated with the chest strap and HR_max was the highest 1-s value achieved during the test. This value was then used to calculate the THR for the experimental trials. Immediately after completion of the GXT, RPE was obtained from participants (Borg and Noble, 1974). Then, approximately 3–5 min later, a 2-mL blood sample was drawn from a superficial forearm vein into a Vacutainer tube containing EDTA (BD Vacutainer, Becton, Dickinson and Co., Franklin Lakes, NJ, United States) for the measurement of blood lactate in duplicate (YSI 2300 STAT Plus, Yellow Spring Instruments, OH, United States). Researchers provided verbal encouragement to participants during all GXTs.

Twenty min following the GXT, participants completed a V ˙ O_2max plateau verification protocol in which they cycled to volitional exhaustion. Those who completed <1 min of the final stage of the initial GXT performed the verification protocol at the final power output achieved during the initial GXT; those who completed ≥1 min of the final stage of the initial GXT performed the verification protocol at a power output 25 W higher than that achieved during the final stage of the initial GXT (Wingo et al., 2005). To be eligible to continue participation in the study all participants had to exhibit a V ˙ O_2max ≥ 20th percentile for cycle ergometer-based testing for their sex and age (American College of Sports Medicine, 2022).

At least 24 h following the control trial, participants returned to the laboratory for the first experimental trial. In addition to the procedures outlined under “all trials,” for the experimental trials, participants measured nude body mass and inserted a flexible rectal thermistor 10 cm beyond the anal sphincter for measurement of T_re. The thermistor was integrated with wireless amplifiers (BioNomadix Wireless SKT Transmitter, Biopac Systems, Inc., Goleta, CA, United States) set to a sampling frequency of 1,000 Hz. T_re and ambient temperature were recorded continuously using a data acquisition system (MP150, Biopac Systems, Inc., Goleta, CA, United States) powered by data analysis software (AcqKnowledge 4.2, Biopac Systems, Inc., Goleta, CA, United States). During the trials based on HR, a member of the research team monitored HR and adjusted the workload to maintain HR within 5 b⋅min^–1 of THR during the entire workout or at 30 W during the recovery intervals if THR was not achievable. During trials based on RPE, the participant adjusted the power output (with the power output concealed) to match the target RPE. The RPE scale was continuously visible to participants throughout the exercise sessions and participants were frequently reminded to adjust resistance to remain at the prescribed intensity. During the HR-based trials, participants were instructed to point to a value on the chart that matched their RPE at the end of the first and fifth work and recovery intervals; the value was verbally confirmed by a member of the research team.

All blood samples taken before, during, and after the experimental trials were drawn from a superficial forearm vein into a Vacutainer containing EDTA for measurement of either lactate concentration, hematocrit (HCT) and hemoglobin (Hb) concentration, or both, as specified in Figure 1. HCT was assessed in triplicate using a microcapillary reader (Model 3201, International Equipment Co., Boston, MA, United States); Hb concentration was assessed in duplicate using a Hb analyzer (HemoPoint H2, EKF Diagnostics, Inc., Boerne, TX, United States). HCT and Hb were then used to calculate plasma volume change (Dill and Costill, 1974). After the last round of HIIT recovery, participants immediately began a GXT in the same manner as during the control trial with no cessation of exercise to determine V ˙ O_2max.

Even though the plateau verification procedure for V ˙ O_2max that was completed during the control trials was not completed in the experimental trials, the V ˙ O₂ values measured after the 4 experimental trials were still referred to as V ˙ O_2max (instead of V ˙ O_2peak). Expressing the values as V ˙ O_2max signified the observed changes were temporary changes in V ˙ O_2max, which is consistent with the nomenclature used in other studies involving cardiovascular drift and V ˙ O_2max (Stone et al., 2021; Lafrenz et al., 2008).

Mean data were generated on the indicated outcome measures. To test the significance of mean differences in power output, a 2-way [condition × time (work intervals 1 and 5)] repeated measures analysis of variance (ANOVA) was used. Power output was also assessed by comparing the change in power output from the first work interval (min 9–11) to the fifth work interval (min 37–40) between the 43_HR and 43_RPE trials using a paired samples t-test. Paired samples t-tests were also used to evaluate the difference in T_re, T ¯ sk , T ¯ b , T_re– T ¯ sk (core-to-skin thermal gradient), and session RPE at the end of the GXT of the 43-min trials.

Baseline data for control and experimental trials were analyzed using a 1-way repeated measures ANOVA. Planned contrasts were performed to compare V ˙ O_2max from each experimental trial to the control trial, using the Bonferroni correction to control for family-wise error rate (α′ = 0.05/number of contrasts). Two-way repeated measures ANOVAs [condition × time (after 15 and after 43 min)] were used to compare V ˙ O_2max and other variables after the GXT after 15 min (1 round of HIIT) to after 43 min (5 rounds of HIIT). To evaluate if V ˙ O_2max decreased by a larger amount depending on the method of exercise prescription (THR or RPE), a paired samples t-test compared changes in V ˙ O_2max (from after 15 min to after 43 min).

For hematological variables, 2-way repeated measures ANOVAs [condition × time (work intervals 1 and 5)] were conducted. For other variables, such as HR, T_re, T ¯ sk , power output, and V ˙ O₂, 2-way repeated measures ANOVAs [condition × time (work intervals 1 and 5) and/or condition × time (recovery intervals 1 and 5)] were conducted. In the event of a significant omnibus test, paired samples t-tests with a Bonferroni-adjusted α level (α′) were used for post hoc comparisons as appropriate. Effect sizes (ES) for paired samples t-tests were calculated using the following formula (Lakens, 2013) for Cohen’s d _av (Cohen, 1988), adjusted for positive bias using Hedges’s correction (g _av ): ES = M e a n   d i f f e r e n c e S D 1 + S D 2 2 × 1 − 3 4 n × 2 − 9 where, SD₁ and SD₂ are the standard deviations of the respective time points or conditions and n is the number of pairs. ES were interpreted as: 0.20 = small, 0.50 = medium, and 0.80 = large, respectively (Caldwell and Cheuvront, 2019; Fritz et al., 2012).

For select variables, the 95% confidence interval (CI) was calculated for the mean difference between conditions (for pairwise comparisons of interest) using a critical t (adjusted, if applicable, to keep the family-wise error rate α at 0.05) in the following formula (Weir and Vincent, 2020): CI = Mean difference  ±   t cv SE d , where t_cv is the critical t value (adjusted for multiple comparisons, if applicable) and SE_d is the standard error of the differences.

For power output, V ˙ O₂, and M–W, the average over the entire interval was used for data analysis; for T_re and T ¯ sk , the average of the final min of the interval was used for data analysis; for HR, both the average over the entire interval and the average of the final min were analyzed. All statistical tests used an α level of 0.05 and analyses were performed using SPSS for Mac v.28.0.0.0 (IBM Corporation, Somers, NY).

Participants were adequately hydrated prior to all trials (mean ± SD, U_SG control = 1.005 ± 0.002, 15_HR = 1.007 ± 0.005, 15_RPE = 1.005 ± 0.004, 43_HR = 1.006 ± 0.006, 43_RPE = 1.006 ± 0.003; p = 0.82). Additionally, pre-exercise body mass was comparable among trials (control = 73.9 ± 7.9 kg, 15_HR = 73.9 ± 7.6 kg, 15_RPE = 74.4 ± 8.8 kg, 43_HR = 74.3 ± 8.9 kg, 43_RPE = 74.0 ± 8.0 kg; p = 0.68). Percent change in body mass from before to after exercise for each experimental trial was greater in the 45-min vs. 15-min trials (15_HR = −0.7% ± 0.4%, 15_RPE = −0.6% ± 0.4%, 43_HR = −1.3% ± 0.8%, and 43_RPE = −1.3% ± 0.8%; p = 0.004 for main effect of time). Additionally, percent change in plasma volume pre-to post-HIIT exercise was not different between 43-min trials (43_HR = −9.0% ± 3.3%, 43_RPE = −9.5% ± 3.9%, p = 0.67).

Baseline T_re was not different among the 4 experimental trials (15_HR = 37.3°C ± 0.3°C, 15_RPE = 37.2°C ± 0.3°C, 43_HR = 37.1°C ± 0.3°C, 43_RPE = 37.1°C ± 0.3°C, p = 0.19). M–W increased over time in 43_RPE and decreased over time in 43_HR. Nonetheless, T_re increased by a comparable amount between experimental conditions so that T_re at the end of the final recovery interval was not different between conditions (MD = 0.2 ± 0.2; 95% CI for MD = −0.05, 0.46; ES = 0.36; Figure 3). T_re was also similar between conditions at the end of the first recovery interval. By the end of the GXT in the 43-min trials, T_re was higher following the RPE trial compared to the HR trial (p = 0.03). Across both experimental conditions, T ¯ sk increased from the end of the first to final recovery interval by ≈ 0.4°C. The core-to-skin thermal gradient (T_re– T ¯ sk ) followed a similar pattern and increased by 0.3°C in both trials. T ¯ b increased by 0.6°C in 43_HR and by 0.8°C in 43_RPE from the end of the first to final recovery interval.

Maximal responses are shown in Table 2 and Figure 4. Planned comparisons between control V ˙ O_2max (3.2 ± 1.2 L⋅min^–1) and V ˙ O_2max after each experimental trial did not reveal any differences [(α′ = 0.0125) p = 0.58 for 15_HR; p = 0.52 for 15_RPE; p = 0.014 for 43_HR; p = 0.014 for 43_RPE]. V ˙ O_2max decreased 15.6% between 15_RPE and 43_RPE (MD = 0.5 ± 0.3 L⋅min^–1; 95% CI for MD = 0.08, 0.87; ES = 0.41; Figure 4), but it was not different over time during the HR-based trials [MD = −0.2 ± 0.2 L⋅min^–1, 95% CI for MD = −0.05, 0.47; ES = 0.16). Furthermore, the change score (15-min value minus 43-min value) in V ˙ O_2max for RPE-based trials was greater than HR-based trials (MD = 0.3 ± 0.3 L⋅min^–1, 95% CI for MD = 0.01, 0.52; ES = 1.13). However, V ˙ O₂ was not different between overtime or between conditions (p = 0.16 for interaction effect). During 43_HR, because V ˙ O_2max and absolute V ˙ O₂ during HIIT exercise did not change over time, relative intensity ( V ˙ O₂ expressed as a percentage of V ˙ O_2max at the specified time point) was maintained from work interval 1 (72% ± 5% V ˙ O_2max) to work interval 5 (69% ± 8% V ˙ O_2max, p = 0.17). Because V ˙ O_2max decreased during 43_RPE and absolute V ˙ O₂ did not change over time, relative intensity increased by 11 percentage units from work interval 1 (70% ± 10% V ˙ O_2max) to work interval 5 (81% ± 11% V ˙ O_2max, p = 0.006; p = 0.002 for interaction effect), and it was 11.5 percentage units higher on average during work interval 5 in 43_RPE versus 43_HR (p = 0.01).

For the 43-min trials, the maximal power output achieved during the GXT was lower compared to the respective 15-min trial. Additionally, the maximal power output during the GXT was 31 W lower in the 43_RPE trial versus the 43_HR trial. T_re and T ¯ b were both 0.2°C higher on average (ES = 0.47 for both) upon completion of the GXT following 43_RPE compared to 43_HR; however, T ¯ sk and the core-to-skin thermal gradient (T_re– T ¯ sk ) were not different between conditions at maximal effort in the 43-min trials.

A limitation of using RPE to prescribe exercise intensity is the different interpretations of the scale. The RPE scale was explained to participants using standardized instructions and they were instructed to complete the work intervals at an RPE of 17 which meant adjusting the resistance up or down to elicit the prescribed RPE. Nonetheless, it appears 1 participant paced themselves and increased power output from the first to final work interval in the 43_RPE trial. This participant preserved their V ˙ O_2max compared to the 15-min trial while the remaining 7 participants experienced an 18% decline (0.6 L min^–1) in V ˙ O_2max on average between the 15-min and 43-min trials. The participant who started slower may have employed teleoanticipation whereby exercise intensity is regulated based on the anticipated endpoint of the exercise session (Ulmer, 1996), although it is unclear why only 1 person may have adopted this strategy. The large range in adjustments during the RPE trial could be partially attributed to the difference in fitness levels of participants; however, this should not have affected the results because of the repeated measures study design. RPE is easy to use for exercise prescription and it can be practical for prescribing HIIT (Buchheit and Laursen, 2013), but the range of work rate adjustments during the work intervals of 43_RPE (−69 W to +19 W) highlights the challenges in using RPE.

Another challenge with using RPE to prescribe intensity during exercise in the heat is the disassociation between the prescribed intensity and the HR response observed during recovery intervals. For instance, participants were instructed to cycle at an RPE of 12 during recovery, which is considered a moderate intensity corresponding to 64%–74% HR_max (Garber et al., 2011). However, based on %HR_max, participants exercised at a vigorous intensity (85% HR_max) during the final recovery interval (Garber et al., 2011). Even during the first interval of recovery, based on HR, participants were at a vigorous intensity (79% HR_max). HR was elevated to such an extent that the %HR_max during the last 3 recovery intervals of 43_RPE were about the same as the work intervals of the 43_HR trial (Figure 2A). Furthermore, 2 participants achieved or surpassed the HR_max observed during the control trial during the submaximal HIIT portion of 43_RPE. Using THR in the heat also proved problematic for prescribing intensity during recovery intervals because most participants were unable to achieve the THR and instead cycled at 30 W.

Despite the limitations of using RPE and THR to prescribe intensity in the present study, both resulted in the exercise intensity being attainable across the varying fitness levels of the participants.

Using target RPE and THR to prescribe HIIT exercise in the heat resulted in considerable declines in work rate during the work intervals, and lower work rates were needed to maintain THR compared to target RPE during the recovery intervals. The higher power output sustained in 43_RPE recovery intervals corresponded to elevated cardiovascular strain during both work and recovery intervals, as well as a greater decline in V ˙ O_2max over time. The non-significant, moderately smaller reduction in work rate from the first to fifth work interval of 43_RPE may have also contributed to the increased cardiovascular strain observed. Although with both methods of exercise prescription reductions in work rate were necessary to maintain the target intensity, all participants (regardless of varying fitness levels; e.g., V ˙ O_2max range: 1.9–5.4 L ⋅ min^‒1) were able to complete the HIIT exercise protocol, which as our pilot testing indicated, would not have been possible if work rate adjustments had not been made. If total energy expenditure is the goal of the exercise session and magnitude of cardiovascular strain is not important, using RPE to prescribe intensity during HIIT exercise may be preferable. Using THR may be preferable if cardiovascular strain is a concern; however, this may limit the training stimulus since larger declines in work rate are necessary.